Mental health care has continued to shift away from extensive inpatient treatment and toward providing more services in community settings. This has led to many calls for managing violence in the community. To do this, however, we must know not only who is at risk for violence, but when these individuals are most likely to be violent. This study proposes to intensively follow a selected group of 125 patients for six months in the community and to provide a detailed picture of the types and patterns of violence in the lives of these patients. Using a recently developed screen, the investigators will identify patients coming into a hospital who are highly likely to be repetitively violent. They will gather information about violent incidents, threats, symptoms, drinking and drug use, treatment involvement, and social relationships through weekly interviews with the patient and a collateral (a person whom the patient has named as knowing about the patient's life). Official records such as arrest records and hospital records will also be obtained. Extensive data about subject selection bias, attrition, and the effects of repeated interviewing will be kept throughout the project, providing valuable information for future investigations of this sort. This study is designed to a) describe the degree to which risk of violence varies over time within patients, and b) test whether changes over time in the risk of violence are predictable from changes in the other measured variables. This study will provide the first data analyses regarding short term changes in risk state for violence.